Jump to content
UltravioletPhotography

Color versus Wavelength


Andrea B.

Recommended Posts

What Color Is It?

by Andrew T. Young

 

Many people incorrectly suppose that there is a one-to-one relation between wavelength and color, or that such a relation exists between spectral energy distribution and color. But even the existence of a unique color (under fixed viewing conditions) for each monochromatic wavelength or each spectral distribution does not imply that a converse relation exists. In fact, it cannot, because color spaces have lower dimensionality than the vector spaces needed to represent spectra; infinitely many spectra necessarily map into the same color.

 

See also Rendering Spectra

 

One of the most difficult tasks in displaying colored images is to render spectra even approximately correctly. The problem is not only that the fully-saturated colors of the spectrum cannot be exactly reproduced in any medium but monochromatic light itself; there is also the problem of mapping the spectral hues into the chosen display medium.

 

....The spectral colors are far outside the gamut of colors that can be reproduced with the standard phosphors.

 

 

**********

 

Similar statement holds for our digicams. And these articles are about "visible" spectra. All problems in representation are compounded for false colours from "invisible" spectra.

 

**********

 

Andrew Young is an astronomer at San Diego State University in San Diego, California.

Link to comment

But this is the first time I've ever encountered a vector space explanation of why the wavelengthcolour mapping is not 1-1. Gotta love that!! :( You can't argue with the dimensionality of the vector spaces: dim3 onto dim2 is for sure many-to-one.

 

....and of course I'm annoyed I didn't think of it myself....oh well.

Link to comment

Hi!

 

>>>But even the existence of a unique color (under fixed viewing conditions) for each monochromatic wavelength or each spectral distribution does not imply that a converse relation exists. In fact, it cannot, because color spaces have lower dimensionality than the vector spaces needed to represent spectra; <<<

 

If I have only one single wavelength then I only have one possible (visible) color (at defined intensity). Backwards if I have a color then I can tell the wavelength if I am sure that it is only one. Going to vectors I compress the space to an area by excluding combinations of vectors outside the area.

 

Am I wrong here?

Link to comment

If I have only one single wavelength then I only have one possible (visible) color (at defined intensity).

 

This is reasonable assuming you can control all the variables which go into making a digital false colour. That is pretty difficult, but we do get somewhat repeatible results with Bayer filtered digicams.

 

Backwards if I have a color then I can tell the wavelength if I am sure that it is only one.

 

If you are sure that is the only wavelength? Well, yes, if you are sitting in a lab with a monochromator under controlled conditions, then you know you only have one wavelength. But in the field you do not know at all "how many" wavelength(s) you have. That's the problem we are discussing - the attempt to determine wavelengths from a false colour foto. You cannot really know what combinations of sine waves produced those false colours.

Link to comment

With lasers, there are light sources at hand, which can produce a light with almost a single wavelength. Assuming our 365nm led would be that good, than the pictures the camera takes would only represent the reflectance e.g. of a flower with respect to that single wavelength. If the picture is not monochrome, then we could assume, that the combination of Bayer-filters and sensor simply might not respond in a linear way to different values (intensity) of the light.

 

We can test this with one of the red or green laser pointer in the visual light (though there, the camera manufactures might have taken counter measures to this). (Which WB to choose? )

Link to comment
Use one of the tools like Raw Digger to see what the actual raw data is and then you don't need to worry about white balance.
Link to comment

Or use a Kelvin WB.

 

I don't tell "I know how it works" but it should be possible.

 

Of course there will be tolerances like in vis photography. But vis photography shows that it is possible. Otherwise a color checker wouldn't work.

 

If I have three light sources with known wavelength all in uv and I shoot three pics with defined settings then I shold be able to create a picture with reproduceable uv false colors.

 

At least it well be possible to find rules to create a pic and of course this rules should be invertible.

 

If it is possible to transfer that to other conditions and cams?

 

Good UV LED's are almost monochromatic. Of course they have internal FL but this is longer in wavelength and one can filter that out.

Link to comment
If you photograph the colour orange in some scene, can you tell me just by looking at the photograph whether the orange colour comes from a combo of Red+Green wavelengths or from a purely spectral single wavelength Orange? If so, how?
Link to comment

No, certainly not.

 

But if I take photo of an orange spot the picture will be orange independent if it's origin.

 

Why should this be impossible in UV. Of course it is in reality but I suppose not because of theory but insufficient euipment.

Link to comment

Yes the picture will be orange. But the question we are trying to answer is which orange has been recorded in the photo? The orange produced by a monochromatic, spectral orange signal? Or an orange produced by a mixed red + green signal? I claim this cannot be determined.

 

Similarly in UV.

If we record UV as a false colour, we do not know if that false colour was produced by a monochromatic signal or by a mixed signal.

 

We already know we can produce standardized false colours in UV with Bayer-filtered digital cameras. We've been doing that for years.

Link to comment

Here, look at the chart in this link which illustrates colour perception in a CIE chart for the human eye.

http://hyperphysics..../colper.html#c1

 

Note that it says that the "combination of light wavelengths to produce a given perceived colour is not unique".

 

Now the same thing holds true for our cameras because the Bayer filters are chosen to produce R, G and B channels in a similar way to how our human R, G and B colour receptors function.

 

So a digital, Bayer-filtered camera cannot be used as a spectrometer in any waveband Vis, IR or UV.

Link to comment

ok, this picture of the color checker was taken with a not modified Olympus OM E1, the Oly 40-150/2.8 lens and a coherent monochromatic laser light and then with PN white balanced (just a click somewhere, doesn't matter much, where exactly).

 

So what is the colour of the laser ? (;-))

 

post-21-0-62824600-1426368171.jpg

 

 

It is red !

Link to comment

I don't know if people don't understand the question - or if they don't understand the physics of light - or if they don't understand colour perception by eyes or cameras. Why does everyone think they can backtrack false colour to a particular UV wavelength? It seems so OBVIOUS to me that this cannot be done.

 

Of course it could be that I do not understand the question or understand the physics of light or understand colour perception by eyes and cameras. This could happen too.

 

But I've gotta go for now. Be back on Monday.

Link to comment

Werner, just out of curiosity, what does a demosaic-ed only version of this photo show? What does the raw data show? Both should show red for a monochromatic light source.

 

What you have illustrated is the fact that the white balance step can alter the raw data. We already know that.

 

What we want to look at is the cases where the RAW data can be the SAME for both a monochromatic light source and a mixed light source. And then we want to ask whether we can determine what the actual light source is simply by looking at the raw data. I claim we cannot.

Link to comment

Andrea, you posted about the same time. Yes, that is what I wanted to show by doing a WB on the above picture.

 

 

So a digital, Bayer-filtered camera cannot be used as a spectrometer in any waveband Vis, IR or UV.

Link to comment

It does not matter, whether you look at the raw data, because you need some kind of calibration. The Raw data are just some kind of electrical signals, amplified some how. And even with a monochrome source of light.

 

 

And this is the raw via PS:

 

post-21-0-28750600-1426369246.jpg

 

 

and the only demosaiced (PN) version:

 

post-21-0-53050300-1426412220.jpg

 

 

 

 

(edits: further pictures included)

Link to comment

You are of course correct that wb does not really matter if the same calibration is applied. :unsure:

But that white balance step is very different across the various converters & editors, so it is probably a good idea to just stick with the raw data for those who want to try playing with this wavelength-to-colour correlation. It makes more sense to to try to correlate to the raw channel information and skip some of the software steps which could just take that data further from what was recorded.

 

Here's an example of raw versus white-balanced data. It is quite interesting for a Visible photo to see how much difference there is. This is a very pink Rosa blanda flower.

 

The raw shot looks like this when demosaic-ed only.

rosaBlandaVisSun_071214seawallMdiME_2799origpn_dem.jpg

 

After applying the camera's "gamma curve" which brightens the midtones, we get this version.

rosaBlandaVisSun_071214seawallMdiME_2799origpn_demGam.jpg

 

And now here is the addition of a proper white balance and profiled colour for this camera, a D600.

rosaBlandaVisSun_071214seawallMdiME_2799origpn.jpg

 

 

 

 

In the UV case the raw shot looks like this when demosaic-ed only.

rosaBlandaUVBaadSB14_070814swFoodMartSwhME_20043origpn_dem.jpg

 

After applying the camera's "gamma curve", we get a little bit blown out.

rosaBlandaUVBaadSB14_070814swFoodMartSwhME_20043origpn_demGam.jpg

 

And now here is the addition of a white balance step on the false colours together with the D600 profiled colour.

You can see how some channel information has been lost from the original raw version.

rosaBlandaUVBaadSB14_070814swFoodMartSwhME_20043origpn_demGamProf.jpg

Link to comment

>I don't know if people don't understand the question - or if they don't understand the physics of light - or if they don't understand colour perception by eyes or cameras. Why does everyone think they can backtrack false colour to a particular UV wavelength? It seems so OBVIOUS to me that this cannot be done.

 

Excuse me please I have to apologize. I still don't understand the question! Excuse me for being the fool but I don't understand that.

 

I don't think that I can backtrack (false) colour to a particular (UV) wavelength. Never in visible and of course never in UV.

 

But the orange dot in a vis picture shows me there was orange in the scene (maybe for example orange light to white surface or white light to orange surface or even red light to a yellow surface possibly producing the same image). So does a certain uv scene not create an according picture?

Link to comment

I'm very sorry, I should have restated everything at the beginning of the thread. There have been so many conversations about this over the last 3 years or so that I'm not really sure anymore who know what. :unsure:

 

********************

 

The Problem:

 

Some folks have photographed UV output from a monochromator using a converted digital camera and lens under a BaaderU filter and collated the results in the form of a wavelength color chart. Nothing wrong with that. It is interesting and does show, for a particular camera, that UV wavelength X is recorded in a certain channel combo X(RGB), that is more simply to say, that UV wavelength X gives a certain perceived colour.

 

However, to map in the reverse direction, as some have wanted to do with these colour charts, is impossible. That is, to use recorded channel data (that is, a certain perceived colour) to infer the reflected wavelength(s) cannot be done.

 

From the physics of light we know that the mapping between visible wavelengths and perceived (reflected) colour is many-to-one. Similarly so, the mapping between UV wavelengths and pereived (reflected) false colour will also be many-to-one. This does presuppose that a standardized method of producing false colours from a particular combo of camera/lens/filter is agreed upon.

 

EDIT: To produce a camera colour chart in the laboratory using a monochromator, one does not use a lens and filter. So I have corrected that error above. Thanks to Klaus for catching this.

Link to comment

Ok, I understand.

 

So it is not possible to tell which wavelength(s) produced the picture.

 

But it should be possible to say that wavelength x could have caused the channel data if there was no other source of light of another wavelength. Knowing that infinite different combination of waves could have caused equal data in the channels.

 

Is that true?

Link to comment

if there was no other source of light of another wavelength.

 

??!!??!!??

 

If you know before you push the shutter that your only source of light is wavelength X, then what else could you possible get except for colour X(RGB) that you already have in your chart. Because that is how you made the chart.

Link to comment
What the camera color chart folks have done is simply to make the first step in creating a "UV false colour space". They have mapped the spectral wavelengths between 300nm - 400nm to a camera's RGB colour. From that they could continue just like for any colour space mapping and create one of the truncated parabolic diagrams just like the basic CIE chromacity diagram. For such a UV colour space the values along the rim would be those spectral values between 300nm - 400nm with the colours on the edges corresponding to what their camera recorded. The interior colours would be combinations of the spectral colours. Any single point on the interior could be mapped from an infinite number of combinations on the rim.
Link to comment

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
×
×
  • Create New...